Generally, a device for recognizing a vehicle is a device for recognizing a vehicle license plate number by acquiring an image of a vehicle and analyzing the image, and exists in various forms, such as an automated enforcement device (for example, speed or signal violation), a traffic information collection device, a device for recognizing vehicle license plate number for security, a camera for recognizing vehicle license plate number on a highway entrance and exit road, and a parking control device, and numbers and types thereof continuously increase.
A vehicle license plate number recognition rate of the device for recognizing vehicle license plate number is in a level of 95 to 98% at maximum, and it is not able to perfectly recognize the vehicle license plate number with 100% due to limitation of an image processing technology.
Actually, it is fundamentally impossible for a recognition algorithm to have a function of perfectly recognizing a vehicle license plate number with 100% due to various noises of the acquired image, image acquiring environment, and damage, contamination, artificial occlusion or the like of a vehicle license plate.
However, unrecognized or misread data of 2 to 5% is generated even in a vehicle license plate number recognition algorithm with the best performance, this is limitation of an original purpose of using the device for recognizing vehicle license plate number, and it is impossible for a current technology to do such a thing or it requires enormous cost and time.
For example, the number of vehicles acquired by a camera of the device for recognizing a vehicle license plate number installed in a city ‘S’ of a metropolitan area is approximately 700,000 vehicles per day, and even if a system operates in an optimal state (a recognition rate of 95 to 98%), that are approximately 14,000 to 35,000 vehicles per day which fail to be recognized.
In such a situation, in case where a search for vehicle license plate numbers are required for one month, there is a problem that an operator has to identify approximately 420,000 to 1,050,000 vehicles with the naked eye.
Since a task of recognizing the images with the naked eye is monotonous, it is easy to lose attention, and even if one person searches 20,000 vehicles per day, it takes about 21 to 50 days. Hence, it is actually impossible for a police officer in charge of investigation to quickly and accurately search for the vehicles in a mastered state. That is, the task of recognizing the images with the naked eye requires considerable time and cost.